When the machine imposes its stereotypes.

AIs learn from the data they are given. If the data is biased, the results will be too. If AI were used in an uncritical way by radio professionals, it could reproduce and amplify stereotypes, including gender stereotypes. This idea raises concerns about side effects: does AI favour certain voices? Does it suppress certain tones or nuances? Human editorial input remains essential to guarantee diversity.

Sub-themes around this idea for programme content: ways to leverage, develop or angle the topic:

For listeners:

• Translation AI: what is lost in the meaning of the original text?

Why do your AI-generated music choices lack surprise?

• Does the algorithm favour certain cultures?
• AI and accents or dialects?


Potential guests: AI researcher, activist defending linguistic minorities, gender equality specialist, etc.

For the radio teams:

• Radio exercise: AI and sound diversity quiz
• What strategies could be put in place to ensure that AI nourishes a wider range of perspectives on air?
• How can AI be used to analyse feedback from Radio listeners and social media in real time, in order to identify and counter violent or discriminatory speech against women and girls?

Useful links: 

- Challenging systematic prejudices: an investigation into bias against women and girls in large language models
- Tables rondes ministérielles sur les bonnes pratiques s’agissant de lever les obstacles existants pour réduire la fracture numérique entre les genres et promouvoir l’éducation à l’ère du numérique aux fins de la réalisation de l’égalité des genres